Welcome to the Hume AI Blog
Published on Aug 3, 2022
Welcome to the Hume AI blog, your source for insights on empathic AI and emotion science. Or, to all you podcast listeners, welcome back! For just about a year now, the blog has been home to The Feelings Lab, our podcast exploring the intersection of emotional intelligence and AI technology that accounts for the complexity of human emotion. Co-hosts Matt and Alan are still in the lab, exploring the depths of emotion science and AI with an ever-growing list of fascinating guests. Check out the latest episode, and subscribe to get new episodes as soon as they drop.
In the meantime, we’re getting ready to make some exciting announcements about what we’ve been working on. So it feels like a good time to start ramping up a steady stream of posts digging into the science, technology, and operations of Hume AI.
Here’s a taste of what’s to come:
Exploring Emotional Intelligence: The Science Behind Empathic AI
Hume AI is an expressive communication research lab and technology company focused on developing AI with empathy. Our mission is to pave the way for artificial intelligence that can teach itself to improve our emotional well-being. The study of human expression is at the heart of this mission and crucial for creating emotionally intelligent voice AI systems.
Great! But … What does it mean to “study human expression”? Good question.
There is an entire academic discipline that studies how people make their feelings and sentiments known to others. To some degree, this happens through language. But so much of how we communicate our feelings isn't about what we say, but how we say it: the nonverbals elements, such as speech prosody (the tune, rhythm, and timbre of speech), vocal bursts (laughs, sighs, groans, umms, uhhs, etc.), and facial expression.
The study of human expression has historically been a non-technical field, and that’s okay—there’s nothing inherently computer this or AI that about the discipline. But when you leave the lab and try to account for real human expression, with its complex patterns of vocal signaling, facial-bodily movements, and emotional language, it turns out that it’s hard to make much sense of all this data without computer programming. To that end, there is a related field, affective computing, that’s focused on training machine learning models to recognize expressive behaviors.
At Hume, we specifically build empathic AI – an advanced form of artificial intelligence designed to recognize, interpret, and respond to human emotions and expressions, enabling more natural and understanding interactions between humans and machines.
Affective computing and empathic AI both seek to enable empathic technology: Applications and systems that are designed to treat human expressive behaviors — like a person’s tone of voice when speaking to Siri or Alexa, patients’ descriptions of symptoms during a telehealth session, or teenagers’ comments on social media — with understanding and empathy. Given that expressions are a significant part of everyday communication, this is a broad goal, with the potential to improve any technology that currently relies on our words alone to understand us.
We at Hume AI believe that understanding expressive behavior is particularly essential to addressing human needs and aligning modern technology with our well-being. Our products give researchers and developers the tools to translate scientific insights into new ways to improve human experience. Our models enable developers to create more sophisticated AI voice generators that can interpret and respond to human emotions.
The science of human expression, like emotions themselves, is a deep, nuanced, and endlessly fascinating thing! The blog will explore the many facets of human expression, from how the field has advanced over time to some of the reasons why human beings act and react the way we do to the news of the world and events in our personal lives.
Advancing Voice AI: Hume's Empathic Technology Platform
Hume AI’s platform provides access to a wide spectrum of models of vocal and nonverbal expression that are complementary to large language models. We offer a new window into the other half of human communication that isn’t explicitly represented in words.
Using new peer-reviewed scientific research, our models are the most accurate and multidimensional to date, enabling developers to capture nuances in expressive behavior with newfound precision. They measure subtle facial movements that express love or admiration, cringes of empathic pain, laughter tinged with awkwardness, and sighs of relief, allowing these signals to be better understood by researchers and used to develop applications that improve people’s well-being.
We train these models using scientific datasets that we collect from hundreds of thousands of consenting participants from around the world. Since these datasets are unique in a lot of ways — larger, more emotionally rich, naturalistic, culturally diverse, and equitable than any data researchers have relied on in the past — we provide them to research teams seeking to train and evaluate unbiased empathic technologies.
We’ve got some exciting updates to share in the coming weeks and months. Can’t say more just yet, but we’re going to share them on the blog. Keep an eye out!
Building with Empathy: AI Voice API and Developer Resources
We’re striving to make our platform as easy to use as it is comprehensive and powerful. Developers can use our Documentation to get up and running on the platform right away. We’ll also be posting tutorials to walk you through key features, and step-by-step How-Tos that get deeper into the power of our datasets and models.
As we expand the platform’s capabilities, we’ll post the latest developer news to the blog, as well.
Join the Conversation about Empathic AI
Follow Hume AI on social media to stay updated on the latest in empathic AI and voice AI technology. Our social channels are also the place to keep tabs on where and when the team is headed to conferences and events — we love connecting with researchers and developers in person!
Take a moment now to follow us on LinkedIn and Twitter, if you haven’t already.
Collaborate on Conversational AI: Let's Build the Future of Emotional Technology
Reach out if you're interested in exploring how emotionally intelligent AI can transform your applications or research. We'd love to discuss how our AI voice API can enhance your conversational AI projects.
You can find us on the social channels listed above, or say hello directly.
Subscribe
Sign up now to get notified of any updates or new articles.
Share article
Recent articles
00/00
We’re introducing Voice Control, a novel interpretability-based method that brings precise control to AI voice customization without the risks of voice cloning. Our tool gives developers control over 10 voice dimensions, labeled “masculine/feminine,” “assertiveness,” “buoyancy,” “confidence,” “enthusiasm,” “nasality,” “relaxedness,” “smoothness,” “tepidity,” and “tightness.” Unlike prompt-based approaches, Voice Control enables continuous adjustments along these dimensions, allowing for precise control and making voice modifications reproducible across sessions.
Hume AI creates emotionally intelligent voice interactions with Claude
Hume AI trained its speech-language foundation model to verbalize Claude responses, powering natural, empathic voice conversations that help developers build trust with users in healthcare, customer service, and consumer applications.
How EverFriends.ai uses empathic AI for eldercare
To truly connect with users and provide a natural, empathic experience, EverFriends.ai needed an AI solution capable of understanding and responding to emotional cues. They found their answer in Hume's Empathic Voice Interface (EVI). EVI merges generative language and voice into a single model trained specifically for emotional intelligence, enabling it to emphasize the right words, laugh or sigh at appropriate times, and much more, guided by language prompting to suit any particular use case.